Ubuntu Linux for Server Networks in Science

1
Ubuntu Linux for Server Networks in Science

• Tony Travis
• Rowett Research Institute
• Aberdeen

2
Tony Travis (RRI, Aberdeen)?

• Job
• Bioinformatician
• Expertise
• Bioinformatics
• WGS sequencing
• Image analysis
• Plant science
• Parallel computing
• Unix/Linux

3
Funding

• Scottish Executive Environment and Rural Affairs
  Department
• European Nutrigenomics Organisation

SEERAD
4
Collaboration

• Computing Science
• Peter Gray
• Alun Preece
• Chris Burnett
• NuGO WP8
• Heiner Boeing
• Ulrich Harttig

German Institute of Human Nutrition,
Potsdam-Rehbrücke
5
Beowulf cluster

• Donald Becker
• Scalable performance
• Commodity hardware
• Private network
• Open source software infrastructure

Grendel
6
EPCC BOBCAT

• Budget-Optimised Beowulf Clusterusing Affordable
  Technology
• Diskless nodes
• Dual ethernet 1 System 2 Application

7
Separate network fabrics

• Cluster servers
• Private
• 192.168.0.0
• 192.168.1.0
• LAN
• 143.234.32.0
• NAT
• Head node
• to LAN/WAN

• Cluster nodes
• Private
• 192.168.0.0
• 192.168.1.0
• NAT
• 192.168.0.0
• To head node

8
Traffic segregation

• System (192.168.0.0) 100Base-T
• PXE boot, DHCP, NFS
• Application (192.168.1.0) Gigabit
• HPC interconnect
• LAN (143.234.32.0) 100Base-T
• Interactive login, HTTP, Web services

9
Distributed process space

• NASA/Scyld
• bproc Linux kernel modification
• Remote process management
• openMOSIX
• Linux kernel-level load balancer
• Automatic process migration
• openSSI
• Single System Image

10
Beowulf operating system

• Ubuntu 6.06.1 LTS
• Kernel level Load-balancing for Linux
• Open source free version of MOSIX
• SSI (Single System Image)?
• Type 2 Beowulf

11
RRI/BioSS Beowulf cluster
Data Filer
Local Filer
sw4
Lab
LAN
RAID
Desktop
Internal users
bobcat
topcat
sw5
sw3
WAN
compute nodes
External users
12
Windows is easier than Unix

• it depends what you want to do...
• Windows is a familiar system
• good desktop environment
• well supported

13
Unix is hard to understand...

• command line
• similar to DOS
• cd
• mkdir
• blastall -p blastx...
• X-Windows
• graphical interface
• similar to Windows

14
Bio-Linux Ubuntu 'biobuntu'

• NEBC (UK) Bio-Linux4 software
• NERC Environmental Bioinformatics Centre
• Ubuntu 6.06.1 LTS operating system
• Open source software based on Debian Linux

15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
Cluster computing

• Embarrassingly parallel
• Distributed processingPVM/MPI

• Multiple instances of CPU intensive tasks running
  on separate processors
• Single instance of a CPU intensive task divided
  up among all available processors

25
Beowulf limitations

• Embarrasingly parallel
• Single CPU performance/memory
• Interconnect bandwidth
• Distributed (MPI)?
• Cluster performance/memory
• Interconnect bandwidth
• openMosix MPI
• Load-balancing assists MPI

26
NuGO Black Box project

• Develop 'lab-scale' server pre-configured with
  base2 and other bioinformatics tools
• Web-based appliance accessed and administered
  using a web browser
• Support remote login via SSH and VNC
• Compatible with NuGO BioMoby
• Encourage data sharing

27
Why a black-box approach?

• Dont need to know how it works to use it
• Deploy a pre-configured Linux server
• Install from live DVD on existing hardware
• Pre-installed on systems supplied by NuGO
• Reduce need for IT support in every lab
• Automatic backup and software updates
• Autonomous system able to discover peer systems
  and cooperate with them

28
(No Transcript)
29
(No Transcript)
30
(No Transcript)
31
Web services

• Enable transparent use of Beowulf cluster as
  compute GRID resource
• Present non-interactive program interface
  Semantic web

32
Services
Model Explorer
Workflow Diagram
33
NuGO-GRID

• Network of NuGO-Linux servers
• Interconnected to create GRID
• Compute GRID
• Load-balance between servers
• Data GRID
• Pool data and share resources
• P2P (Peer-to-peer)
• Local control of resource sharing

34
Current status of NuGO-GRID
35
SSH (Secure Shell)?

• Used to tunnel the insecure database port 3306
  through encrypted channel
• Encrypts data in transit and allows the database
  server port to remain hidden

Encrypted data
Database (port 3306)?
Firewall (port 22)?
Client (port 3306)?
36
Remote administration via SSH
nugo_at_nbx2 sudo nbx-keyscan nbx1.nugo.org
SSH-2.0-OpenSSH_4.2p1 Debian-7ubuntu3.2
nbx2.nugo.org SSH-2.0-OpenSSH_4.2p1
Debian-7ubuntu3.2 getaddrinfo nbx3.nugo.org Name
or service not known getaddrinfo nbx4.nugo.org
Name or service not known nbx5.nugo.org
SSH-2.0-OpenSSH_4.2p1 Debian-7ubuntu3.2 read
(nbx6.nugo.org) No route to host nbx7.nugo.org
SSH-2.0-OpenSSH_4.2p1 Debian-7ubuntu3.2
nbx8.nugo.org SSH-2.0-OpenSSH_4.2p1
Debian-7ubuntu3.2 nbx9.nugo.org
SSH-2.0-OpenSSH_4.2p1 Debian-7ubuntu3.2
nbx10.nugo.org SSH-2.0-OpenSSH_4.2p1
Debian-7ubuntu3.2 getaddrinfo nbx11.nugo.org
Name or service not known nbx13.nugo.org
SSH-2.0-OpenSSH_4.2p1 Debian-7ubuntu3.2
nbx14.nugo.org SSH-2.0-OpenSSH_4.2p1
Debian-7ubuntu3.2 nbx15.nugo.org
SSH-2.0-OpenSSH_4.2p1 Debian-7ubuntu3.2
nbx16.nugo.org SSH-2.0-OpenSSH_4.2p1
Debian-7ubuntu3.2 nbx17.nugo.org
SSH-2.0-OpenSSH_4.2p1 Debian-7ubuntu3.2
nbx18.nugo.org SSH-2.0-OpenSSH_4.2p1
Debian-7ubuntu3.2 nbx19.nugo.org
SSH-2.0-OpenSSH_4.2p1 Debian-7ubuntu3.2
nbx21.nugo.org SSH-2.0-OpenSSH_4.2p1
Debian-7ubuntu3 nbx22.nugo.org
SSH-2.0-OpenSSH_4.2p1 Debian-7ubuntu3
nbx23.nugo.org SSH-2.0-OpenSSH_4.2p1
Debian-7ubuntu3 getaddrinfo nbx24.nugo.org Name
or service not known nbx25.nugo.org
SSH-2.0-OpenSSH_4.2p1 Debian-7ubuntu3.2
nbx26.nugo.org SSH-2.0-OpenSSH_4.2p1
Debian-7ubuntu3.2
37
Dancer's shell / distributed shell

• Used to execute remote commands
• sequentially
• dsh -Mg up -- uptime
• concurrently
• dsh -cMg up -- uptime
• System administration
• local and remote NBX managers

38
IT policy at NuGO partners

• Limited access requested
• Port 22 (SSH) and 80 (HTTP) open
• Tunnel insecure protocols via SSH
• Client PC requirements modest
• Java enabled web browser
• Optional installation of Windows clients
• Remote admin of NBX's by NuGO

39
What else is needed ?

• Transparent p2p exchange of base2 data objects
  between NBX's via BioMoby
• Integrate NuGO R-server and RRI/BioSS Beowulf
  cluster into NuGO BioMoby
• Execute base2 plugins on remote compute servers
  (large memory SMP Beowulf)?
• Establish NBX help desk and training
• Get more feedback from Focus teams

40
In conclusion

• Ubuntu is Debian
• better presented
• better supported
• 'biobuntu'
• Ubuntu 6.06 LTS
• NEBC bio-linux

http//bioinformatics.rri.sari.ac.uk/biobuntu